panic: Bio already on queue bp=0xc1c69210

This is with quite recent -CURRENT, cvsupped yesterday. Machine locks  
after
panic, so no  coredump. Stack trace, written down by hand:

kdb_backtrace()
panic ()
g_io_request ()
g_io_deliver()
g_bde_contribute ()
g_bde_work_done ()
g_bde_start2()
g_bde_start1 ()
g_io_shedule_down ()
g_down_procbody ()
fork_exit ()
fork_trampoline()

How to reproduce:

mdconfig -a  -f  md.img -u 0
gbde attach md0 -l md.lock
mount /dev/md0.bde /mnt
mv foo/ /mnt

This panic is semi-repeatable, (un)fortunately it does'nt crash every time.
Catalog  foo contains lot of small files (at least hundred or so).